comfyui-lrw-nodes — ComfyUI LRW Nodes¶

Development line: project:comfyui-lrw-nodes · thread comfyui-lrw-nodes
Last event: 2026-06-09 · 1 dated since 2026-06-09 · Researched: 2026-09-05 · confidence: medium

What it is¶

ComfyUI-LRW-Nodes provides custom nodes for ComfyUI to guide WAN2.2 First–Last Frame intermediate motion without replacing endpoint conditioning.

• Geodesic and SLERP latent interpolation.
• Latent distance, curvature, and parallel-transport experiments.
• WAN2.2 geodesic keyframe blending into video latents before KSampler.

Requires latent-riemannian-world >= 0.3.0, torch >= 2.4, and extra WAN and video nodes for the supplied workflow. Use it as an experimental soft guide and compare against a fixed-seed direct-FLF baseline; it does not replace WAN2.2 endpoint continuity.

Development line¶

• 2026-06-09 — ComfyUI LRW Nodes repository and workflow examples were documented. A dated message on 2026-06-09 linked ComfyUI LRW Nodes to its GitHub repository and example-workflows directory. This confirms people referenced the project and workflows on that date. It shows no specific release, feature change, or repository state.

What changed¶

• 2026-06-07 — Initial public rollout targeted WAN2.2 First–Last Frame. LRW endpoint latents produced a geodesic guide blended before KSampler. A same-day registration and import-path fix followed.
• 2026-06-09 — The repository and its workflow directory went public. We have no release version or dated changelog for this step.

How to use this¶

As of 2026-06-09, evaluate ComfyUI LRW Nodes through the repository and example workflows.

1. Install through ComfyUI Manager by searching comfyui-lrw-nodes, or clone the repository into ComfyUI/custom_nodes, run pip install -r requirements.txt, and restart ComfyUI. — https://github.com/lajjadred/comfyui-lrw-nodes
2. For WAN2.2, install workflow dependencies: ComfyUI-GGUF, WAN First–Last Frame support, and ComfyUI-VideoHelperSuite; KJNodes is optional. — https://github.com/lajjadred/comfyui-lrw-nodes
3. Load the practical workflow. Feed first and last images to direct WanFirstLastFrameToVideo, then VAE-encode both endpoint images for LRW. — https://github.com/lajjadred/comfyui-lrw-nodes/tree/main/examples/workflows
4. Create geodesic keyframes, pass one through LRW_WanLatentGuideBlend, sample with KSampler, and decode/combine the video. — https://github.com/lajjadred/comfyui-lrw-nodes

Best practices¶

• Keep direct WAN2.2 FLF as the continuity path. Blend the LRW guide into its video latent instead of replacing target last-frame conditioning. — https://github.com/lajjadred/comfyui-lrw-nodes
• Start with blend_strength=0.15, time_schedule=middle_focus, and normalize_guide=true; the documented range is 0.10–0.20. — https://github.com/lajjadred/comfyui-lrw-nodes
• If nodes show as unknown or fail to import, keep one installation only, update or reclone it, restart ComfyUI fully, and inspect the startup log around this node pack. — https://www.reddit.com/r/comfyui/comments/1tz4qnu/i_built_comfyui_nodes_that_use_riemannian/

Superseded by this¶

• 2026-06-07 — Stale __init__.py registration importing LRW_LatentKeyframePicker from wan_nodes.py is obsolete. Update the node pack and restart ComfyUI.
• 2026-06-07 — Guidance that treats LRW as a direct quality upgrade or replacement for WAN2.2 FLF end-frame conditioning is obsolete. Its documented role is a soft pre-KSampler latent guide.

Still unknown¶

• No dated first-party release, tag, or changelog establishes the exact repository revision on 2026-06-09.
• The README lists four example workflows, while the linked workflow directory shows only wan22_practical_direct_flf_plus_lrw_guided_blend.json. Baseline, image-interpolation, and style-transfer JSON files are unverified.
• GitHub reports ten commits, but their dated history could not be retrieved during this check.

Sources¶

Agent brief¶

• Subject: project:comfyui-lrw-nodes, thread comfyui-lrw-nodes, 1 dated events 2026-06-09 → 2026-06-09.
• Practical note: As of 2026-06-09, practitioners can use the linked project repository and example workflows as the starting point for evaluating or adopting ComfyUI LRW Nodes.
• Confidence: medium. Dated supersedes above are the authority for what is obsolete.